Real-time-GesRec
Real-time-GesRec copied to clipboard
Result of online_test on egogesture
Hello! I debug the code and run the online_test.py on egogesture with your pretrained model successfully. However the final result is poor.
report/egogesture_resnext_101_Depth_32.pth, resnext, 1.0, Depth, 0.7691380349608198
Here is my parameter:
python online_test.py
--root_path /Lun4/fdh/Real-time-GesRec
--video_path ../egogesture-dataset/image
--annotation_path annotation_EgoGesture/egogestureall.json
--resume_path_det report/egogesture_resnetl_10_Depth_8.pth
--resume_path_clf report/egogesture_resnext_101_Depth_32.pth
--result_path results
--dataset egogesture
--sample_duration_det 8
--sample_duration_clf 32
--model_det resnetl
--model_clf resnext
--resnext_cardinality 32
--downsample 1
--model_depth_det 10
--width_mult_det 0.5
--model_depth_clf 101
--resnet_shortcut_det A
--resnet_shortcut_clf B
--batch_size 1
--n_classes_det 2
--n_finetune_classes_det 2
--n_classes_clf 83
--n_finetune_classes_clf 83
--n_threads 16
--checkpoint 1
--modality_det Depth
--modality_clf Depth
--n_val_samples 1
--train_crop random
--test_subset test
--det_strategy median
--det_queue_size 4
--det_counter 2
--clf_strategy median
--clf_queue_size 32
--clf_threshold_pre 1.0
--clf_threshold_final 0.15
--stride_len 1 \
Is there anything wrong?
Hi, could you also provide pretrained paths for det and clf which is the same as resume path. In the first look, I do not see a problem but this may be because of a bug in loading the model.
Hi, could you also provide pretrained paths for det and clf which is the same as resume path. In the first look, I do not see a problem but this may be because of a bug in loading the model.
Hello,but the online_test.py doesn't contain the model loading path of pretrained path. And the model loading of resume_path is successfully. Maybe some parameters should be modified? Such as det_counter?
@ahmetgunduz I also debug the code and run the online_test.py on egogesture with your pretrained model successfully. However the final result is poor.The same as @kinfeparty
@ahmetgunduz @kinfeparty I have poor result when evaluate testing dataset. Did you resolve this problem ?
@ahmetgunduz I use jester_resnext_101_RGB_32.pth
@kinfeparty @ahmetgunduz @zhangyuan1994511
same here: report/egogesture_resnext_101_Depth_32.pth, resnext, 1.0, Depth, 0.7691380349608198
Did anyone found any bugs?